Beyond Summaries: Structure-Aware Labeling of Code Changes Enables Prioritized LLM-Assisted Code Review

arXiv·medium signal

Proposes structure-aware labeling of code patches that identifies change types (renames, moves, logic modifications) to enable prioritization, filtering, and automation in LLM-assisted code review. Existing approaches treat patches as flat text; structured labeling lets reviewers skip mechanical changes and focus on logic-altering diffs. Directly applicable to AI-assisted development workflows.